|
Abstract:
|
The field of statistics has become one of the mathematical foundations in detection and attribution studies, which is especially suitable for assessing uncertainties in climate change studies. The classical paradigm is to infer regression coefficients to quantify expected climate response patterns to different external forcings. To take into account of climate model errors, we explore two different approaches. The first one is based on the idea of finding statistical summaries, that, while bringing relevant information, are robust to multi-error errors. In particular, summaries based on relative rankings of records will be studied and their links to generalized extreme value (GEV) distributions developed. The second approach is based on error-in-variable (EIV) models. As the Gaussian hypothesis is not appropriate to handle extremes such as annual maxima, we discuss how to estimate relevant posterior probabilities with different EIV models based on GEVs. Two proposed approaches are applied to global climate model outputs from the Coupled Model Intercomparison Project to attribute anthropogenic forcings on annual maxima of precipitation and/or daily temperatures.
|
